Face super-resolution method based on frequency decomposition multi-attention mechanism

Abstract

The invention belongs to the field of computer vision and image processing, and relates to a face super-resolution method based on a frequency decomposition multi-attention mechanism. The method comprises the following steps: performing frequency decomposition on an input low-resolution face image by utilizing the reversible property of wavelet transform and inverse transform thereof, constructing a basic module by adopting different kernel convolution aiming at the characteristics of different frequencies, integrating the features of different receptive fields adaptively, and processing the features of different frequencies by using residual attention modules including pixel, space and channel attention mechanisms, wherein low-frequency part textures adopt attention with less calculated amount, and high-frequency part adopts more residual attention modules; applying more networks to a high-frequency part while the calculated amount is kept, extracting and feeding back key points by utilizing a pre-trained face key point extraction network, enhancing contour features, and enhancing the texture features by utilizing a generated resistance network.

Description

technical field [0001] The invention belongs to the fields of computer vision and image processing, and relates to a face super-resolution method based on a frequency decomposition multi-attention mechanism. Background technique [0002] The size of the image resolution is directly related to the quality of the image, which will have a great impact on high-level tasks including detection and recognition. Higher resolution means more detailed information and greater application potential. However, in the actual image acquisition process, due to the limitation of imaging equipment itself, the influence of environmental factors, the influence of storage media and network bandwidth, it is impossible to directly acquire high-quality images. With the rapid development of computer vision technology, especially deep learning, there are more and more image enhancement methods. Super-resolution technology is an effective means to improve image quality, which can effectively improve im...

AI Technical Summary

Problems solved by technology

[0004] In order to solve the above-mentioned technical problems existing in the prior art, the present invention proposes a face super-resolution method based on the frequency decomposition multi-attention mechanism, under the condition that the complexity of the network is not increased and the consistency of the low-frequency structure is guaranteed, the The network pays more attention to the high-frequency part, uses the Haar wavelet transform to decompose the original input image into four parts with different frequencies, and enhances the high and low frequency features through the network respectively, and it is difficult to estimate the key points of the face directly from the LR input. The key point information of the face is extracted through the super-resolution output result map and fed back to the low-frequency part of the backbone network to further improve the face information. The specific technical scheme is as follows:

Images